This paper presents an improved new method with differential evolution and the cubic spline approach is proposed to retrieve sea level height based on GNSS SNR observations from a single geodetic receiver. Considering the B-spline function is unstable at the beginning or end, and the feature that B-spline functions do not pass through nodes may introduce errors. Thus, the cubic spline is applied to the retrieval process and accounts for a continuous and smooth in sea level retrieval time series. Besides, the biases caused by tropospheric delay and dynamic sea level are considered and corrected. Testing data from two stations with different tidal range and the final solution agrees well with measurements from co-located tide gauges, reaching the RMSE of 3.67 cm at Friday Harbor, Washington, and 1.36 cm at Onsala, Sweden. Comparison of the nonlinear least squares, this method leads to a clear increase in precision of the sea level retrievals within 50%. Additionally, referring to the result of Purnell et al. (2020) and the IAG inter-comparison campaign, the results of this paper show more potential.